EP1749278A2 - Procede de traitement de donnees d'image, par réduction de bruit d'image, et camera integrant des moyens de mise en oeuvre du procede - Google Patents
Procede de traitement de donnees d'image, par réduction de bruit d'image, et camera integrant des moyens de mise en oeuvre du procedeInfo
- Publication number
- EP1749278A2 EP1749278A2 EP05757055A EP05757055A EP1749278A2 EP 1749278 A2 EP1749278 A2 EP 1749278A2 EP 05757055 A EP05757055 A EP 05757055A EP 05757055 A EP05757055 A EP 05757055A EP 1749278 A2 EP1749278 A2 EP 1749278A2
- Authority
- EP
- European Patent Office
- Prior art keywords
- image
- points
- window
- noise
- values
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20076—Probabilistic image processing
Definitions
- Image data processing method by reducing image noise, and camera integrating means for implementing the method
- the present invention relates to the processing of image data by noise reduction.
- an image appears as a table of values, generally rectangular.
- a temporal sequence of images such as a film
- it is a series of such tables, over time.
- a “time pixel” for a point belonging to a sequence of images over time, for example a point which evolves from one image to another in a film. It will therefore be understood that a “temporal pixel” is associated with a temporal coordinate, additional compared to a conventional pixel.
- a sample for the association of a point and its value.
- Each point is the result of a measurement, generally made by a matrix or a strip of light sensors, such as CCD (for “Couple Charge Device”).
- CCD for “Couple Charge Device”.
- a point corresponds to a small square of the CCD matrix in which the number of arriving photons is account.
- the arrival of photons follows a random process introducing fluctuations around an average value.
- each sensor can itself produce a "dark noise" which is superimposed on the photon count.
- the noise can depend, at each point, on the gray level of this point. This is how there is generally more noise in the bright parts of the image, even if the signal to noise ratio is better there.
- signal / noise ratio t This parameter designates the proportion of a gray or color level that can come from noise. For example, a signal-to-noise ratio of 100 corresponds to fluctuations of one hundredth of the "true value" of the gray level. For images usual digital, the gray or color level varies between 0 (black) and 255 (white). Noise becomes sensitive as soon as it exceeds an average amplitude, or standard deviation, of 4 or 5, which corresponds to a signal / noise ratio of around 50. It is very useful to be able to eliminate all kinds of noise from the images, because the noise interferes with the vision of the images and hides details. Typically, a noise-free image appears sharper.
- CCD sensors are itself dictated in part by the requirement for noise reduction. If we knew how to effectively suppress noise, we could build smaller sensors by applying noise reduction processing to the signals received. We could thus build smaller cameras, with fewer CCD sensors, but providing the same number of pixels as existing ones. We could also design cameras with the same characteristics, but at higher resolution, or identical cameras, but usable with a shorter exposure time.
- noise reduction methods described in particular in documents ' US-6,681,054 and US-6,539,125 use a weighting based on the spatial distances between a point to be treated and current points surrounding this point to be treated, to calculate an average to be assigned to the values of the point to be treated. Consequently, if the image locally presents a wide variety of nuances, after processing by noise reduction based on a weighting with respect to a spatial distance, this variety of nuances is no longer found.
- the present invention improves the situation.
- a method for processing image data by reducing image noise, comprising the following steps: a) obtaining points of the image with respective values associated with the points, b) associating with a image reference point, a learning zone, and c) assign to the reference point new values obtained by a weighted statistical estimate, of weighted average or weighted median type, using the values of the points included in zone d 'learning.
- the process within the meaning of the invention more precisely comprises the following steps: cl) for each current point of the learning zone, evaluate a distance characterizing a resemblance between: - the values of the points in a first window of the image, centered on the reference point, on the one hand, and - the values of the points in a second window, in the same format as the first window and centered on the current point, on the other hand, c2) and repeat step cl) for all the points of the learning zone as points successive currents using the distances obtained for all the current points to calculate the weights used in the weighted statistical estimate of step c).
- the above-mentioned step c) is applied to all the points of the image as successive reference points, in order to process the entire image overall.
- the learning zone can correspond to the entire image to be processed.
- the learning zone belongs to one or more model images, different from the image to be processed.
- the points can be pixels for a two-dimensional image, voxels for a three-dimensional image, or even time pixels when the image to be processed is a film.
- these points are ultimately successive samples.
- step a) of the method within the meaning of the invention the points of the image are acquired from one or more light sensors of given surface, by imposing on this or these sensors, a time d exposure to light, per unit area, predetermined. It is indicated that, generally, a decrease in exposure time leads to an increase in noise.
- steps cl), c2) and c) offers a reduction in noise by a factor K, a reduction, appreciably by a factor K 2 , of the exposure times of the sensor (s) is authorized. , to operate at a substantially constant signal-to-noise ratio.
- the number of sensors per unit area can be advantageously increased, in order to increase, by a factor of K 2 , the resolution of l image acquired and processed.
- the present invention also relates to a camera, equipped with one or more sensors and comprising means for controlling the exposure time of the sensors for the implementation of the above process.
- a processing unit adequately programmed to apply the processing method within the meaning of the invention to the signals acquired by the camera's sensor or sensors.
- this processing unit comprises a memory suitable for storing a computer program product comprising instructions for the implementation of all or part of the steps of the above method.
- the present invention also relates to such a computer program product, intended to be stored in a memory of a processing unit of the aforementioned type, or else on a removable memory medium, such as a CD-ROM or a floppy disk, intended for. cooperate with a reader of the processing unit.
- FIG. 1 very schematically illustrates the elements of a camera involved in the implementation of the processing within the meaning of the invention
- FIG. 2 schematically represents the format of the windows to be compared for the evaluation of the distance, and, from there, of the aforementioned weighting
- FIG. 3 schematically represents a succession of windows of respective increasing sizes used for the evaluation of the aforementioned distance, in an embodiment described below
- - Figure 4 summarizes the main steps of the method according to a preferred embodiment
- - Figure 5 shows additional steps of the above method, in a preferred embodiment.
- the camera defined above comprises a matrix of CCD sensors receiving a light flux from an OBJ objective, the two elements being for example housed in a black box (not shown) . It also includes an acquisition card 1 in particular for converting the signals from CCD sensors into digital samples, typically pixels, which are temporarily stored in a working memory 2 to be processed by a processor 3.
- a digital image is made up of pixels which can be assimilated to points on a grid, provided with a gray value or with color levels.
- a memory 4 for example a read only memory, stores the instructions of a computer program product for the implementation of the method according to the invention.
- the processed pixels can then be transmitted by an interface 5 which, in the example described, is a graphical interface for displaying the processed image on an ECR display screen.
- the interface 5 can be a communication interface for transmitting the processed pixels to a remote entity. It can also be an interface to a storage unit, for example for storing on a memory medium, for later retrieving the pixels. acquired and processed by the process within the meaning of the invention. Finally, it can be an interface to a paper printing unit.
- the samples acquired are transmitted directly to a remote entity comprising a processing unit to apply the noise reduction processing within the meaning of the invention to these samples, rather than having the samples processed by the camera itself.
- FIG. 2 shows a two-dimensional image.
- the image can be three-dimensional and include voxels.
- the image can also be a film and include temporal pixels, without these various distinctions affecting the principles of the invention, as will be seen.
- I the value of the color (or of the gray level for a black and white image) associated with the Pref pixel.
- This value is either an integer or real number (for a grayscale image) or a triplet in the case of RGB standard color images (for red-green-blue), or possibly a tuple in the case of a multispectral image, without these various distinctions still affecting the principles of the invention.
- the objective is to remove the degradations while preserving the main characteristics of the image, as well as the details of small dimensions and the textures.
- the processing within the meaning of the invention makes no assumption about the nature of the noise or of the image.
- Knowledge of the type of noise will however make it possible to specify and adapt several of the elements of the process such as the type of statistical calculation to be carried out (weighted average or weighted median for example), or even the size to be fixed for the windows of resemblance.
- the processing is nevertheless based on the fact that all the usual images have a high degree of redundancy. We then consider that there are around each pixel pixels which are very similar to it.
- the requirement of resemblance is preferred over the criterion of spatial proximity.
- FIG. 2 there is shown a plurality of points PC, PC, ... belonging to a learning zone ZA which can be located in the image to be processed IM or, sometimes more advantageously, in d ' other images considered as models.
- a reference point Pref to be processed and which necessarily belongs to the IM image.
- the window f2 including the point PC includes MOA image patterns in the form of oblique zebra lines, while the windows fl and f'2 respectively including the reference point Pref and the other current point PC are both "white".
- the processing within the meaning of the present invention will allocate a greater distance d to the window f2 than the distance d 'which is allocated to the window f'2, even if, however, the point PC is closer, in distance, to the Pref reference point than the PC point.
- window centered in one pixel is used to mean all the pixels which are at a predetermined distance from the central pixel. It is indicated that these windows are preferably square, here. However, their shape can also be rectangular, oval, or other.
- a criterion of distance between windows is established, this distance being evaluated for example on the basis of a weighted sum of squares of the differences in point values, between the first window and the second window. Any other distance, norm, or measure evaluating the resemblance of the two windows is also possible.
- the resemblance between two windows can be estimated preferentially based on the sum of the squares of the differences of gray levels or color in each window • pixel, as discussed in detail below.
- the weighted average used to calculate the value assigned to the restored pixel of I (Pref) is therefore an average where the similar windows count a lot, while the non-similar windows count little or do not count at all.
- FIG. 4 summarizes these steps.
- a reference pixel Pref to be processed, is designated in the image.
- a window of predetermined size f 1 is associated with it, in step 42.
- the designation of the reference pixel to be processed also makes it possible to associate therewith a learning zone ZA and to designate, consequently, the current pixels PC, PC, ..., belonging to this area of learning ZA and on the basis of which the averaging will be carried out.
- step 44 the method continues with the construction of the second windows f2, f'2, ... associated with the current pixels PC, PC, etc.
- step 45 respective distances d, d ′, etc. are evaluated, according to a similarity criterion, between the window fl and the window f2, between the window fl and the window f'2, etc.
- a weighted average MOY is evaluated as a function of the pixel values val (PC), val (PC), ..., of the current points PC, PC, ..., as well as the respective distances d, d ', ... calculated in step 45.
- MOY val (PC) / D + val (PC) / D' + ...
- D, D ', ... are values that vary as the distances d, d ', ...
- step 47 the value MOY is assigned to the reference pixel Pref and the new value val (Pref) is preferably stored in memory for later display of the image, or other (step 50).
- step 48 a test checks whether there is another pixel of the image to be processed, in which case this new pixel is designated as a reference pixel in step 41, which is implemented again, with the following steps 42 to 50. Otherwise, the process stops at step 49.
- Each pixel p is associated with a "learning zone centered at p" ZA, defined as a window A (p, N) of any shape (square, circular, or other), centered at p and of fixed size (2N +1) (2N + 1).
- the new value I r es (p) associated with p is written as a weighted average of the values I (q) of the pixels q which belong to the learning zone A (p, N), according to a relation of the type: Iresip ) ⁇ (D where the weights (p, q) vary inversely with the distance between the pixels p and q, according to a criterion of resemblance between respective associated windows.
- the resemblance between pixels will depend on the resemblance of color values (or gray levels) between the W (p, M) and W (q, M) windows.
- the weight w (p, q) is all the greater when the pixels in (p, M) and W (q, M) are more similar.
- the weight w (Pref, PC) is large because the resemblance between the windows fl and f'2 including Pref and PC respectively is large.
- the weight w (Pref, PC) is small because the windows of resemblance are different, the window f2 including stripes as indicated above.
- - (p, i) (j) are the (2i + l) 2 values of the points of the window of size (2i + l) x (2i + l) and centered on the point p
- - W (q, i ) (j) are the (2i + l) 2 values of the points of the window of size (2i + l) x (2i + l) and centered on the point q.
- two windows fl and f2 of the same size (2M + l) x (2M + l) and denoted W (p, M) and W (q, M), are similar if and only if their sub-windows (p, i) and W (q, i), of size (2i + l) x (2i + l), with l ⁇ i ⁇ M, are also.
- the window fl centered on the reference point Pref, as in Figure 2, includes a plurality of sub-windows (W (p, l), W (p, 2), (p , 3), ..., W (p, i), ..., W (p, M)) nested one inside the other, centered on the point Pref and of respective increasing sizes (3x3, 5x5, 7x7, ..., (2i + l) x (2i + l), ..., (2M + l) x (2M + l)), from (p, l) to W (p, M).
- the window f2 centered on the current point PC, as in FIG. 2, includes a plurality of sub-windows (W (q, l), (q, 2), W (q, 3), .. ., (q, i), ..., (q, M)) nested one inside the other, centered on the current point PC and of respective increasing sizes (3x3, 5x5, 7x7, ..., (2i + l) x (2i + l), ..., (2M + 1) x (2M + 1)), from (q, l) to (q, M).
- the windows fl and f2, like the sub-windows W (p, i) and W (q, i) (for all the i from 1 to M), are of the same format and, in particular, of the same respective sizes.
- - u (p, i) (j) are the (2i + l) 2 vector values of coordinates u of the points of a current sub-window, of size (2i + l) x (2i + l), included in the first window fl and centered on the reference point p, and
- the weighting assigned to a current point q decreases with the distance between the values associated with this point q and the values associated with the reference point p.
- - h is a parameter indicative of a degree of filtering of the image.
- Z (p) is a constant such that the sum of the weights w (p, q) is equal to 1.
- Z (p) is therefore defined simply from the relation: Z (p) ⁇ YexO (- d [w (p, M ⁇ , W ⁇ q, MW) (5) qeA (p, N)
- This process therefore involves three parameters, which can be set for digital images suitable for applications for the general public as follows.
- the first parameter is the size of the learning zone (2N + 1) x (2N + 1). This area must be large enough to take into account the statistical properties of the image. Tests have shown that the size of the learning zone, in number of points, is preferably greater than 20 ⁇ 20. Typically, an area of 21 ⁇ 21 pixels is sufficient to guarantee proper learning.
- the second parameter is the size of the resemblance window (2M + 1) x (2M + 1).
- a window that is too small can lead to a too local comparison.
- a window that is too large can have the consequence that no window is really similar.
- the size of the first fl and second f2 windows is greater than 10x10 and preferably greater than or equal to 15x15.
- a resemblance window size fl and f2 of 15x15 is optimal for a Gaussian noise of standard deviation 20.
- a window size of 11x11 already provides very good results. This size is indicated for a grayscale image. For a color image, the size of the resemblance window can be reduced to 7x7 and for weak noise up to 3x3.
- the third parameter is the h parameter which indicates the degree of filtering of the image.
- This parameter is preferably fixed at a value close to 3. However, this value can be chosen smaller if the image is slightly degraded.
- the method applied to a two-dimensional image with the values of N, M and h above makes it possible to increase the signal / noise ratio by a ratio greater than 2.
- This measurement can be verified by taking an image with little noise, and by then adding an artificial noise to it.
- the performance of the treatment is measured exactly by comparing the average relative error of the restored image with the noise-free original, as well as the relative error introduced by the noise. We then systematically observe a multiplication by a factor of more two of the signal-to-noise ratio.
- CCD sensors can be used four times smaller (2 2 ) without increasing noise, as indicated above.
- the computer program product within the meaning of the invention can be stored in a memory of a photographic development apparatus, a digital camera, or a device for restoring digital images, but the application of the invention can also influence the design of the sensors of any new camera or photographic device by allowing a signal / noise ratio reduced by a factor greater than four.
- the size in pixels of the resemblance window therefore varies between 3x3 and 15x15 in the zoomed image.
- the 3x3 size in a color image with a noise of standard deviation lower or close to 13 allows to restore the finest details.
- the method within the meaning of the invention comprises the following overall steps, after obtaining the raw image IM in step 51: - in step 52 , we enlarge the image to be processed IM by a digital zoom operation, - the processing TR is applied within the meaning of the invention and preferably corresponding to steps 41 to 50 of FIG. 4, and - in step 55, the image processed 1st is reduced in the opposite direction; ' P has a zoom out operation, to restore the processed image I res -
- step 54 adapt the format M, in particular in number of points, of the resemblance windows as a function of the estimated noise level
- the processing is more efficient also if the value of the restored pixels depends not only on the average of the values of the resemblance windows, but also on the variance of these values, which can be estimated by a process known per se.
- the noise reduction processing within the meaning of the invention generally consists of replacing the value in each pixel of the image by a weighted average of all the values of the pixels of the image.
- the weighting is made in such a way that a window that looks a lot like the pixel-centered window contributes a lot to the average, while a window that doesn't look like that hardly contributes.
- the resemblance between two windows of the same format is evaluated on the basis of a function of the differences in values between the respective pixels of the two windows (for example, a quadratic mean of these differences), or any other standard, difference or distance measuring more precisely the resemblance of the two windows.
- This process makes it possible to use sensors with internal noise or photonic noise more than four times higher, and to eliminate the noise created by various restoration operations (typically clearing, extension of the gray scale for underexposed photographs , or others) .
- weighted average can be replaced by other statistical estimators such as median or weighted median depending on the type of noise considered.
- calculation weights in the weighted average or the weighted median may depend, in addition to the distance based on a similarity criterion as described above, other statistical parameters estimated globally in the image and on each similarity window, such as, for example, the variance of each window or the estimated variance of noise in the image.
- the method described above is suitable for an image of arbitrary size, both in color and in black and white, for a film or a three-dimensional image, whatever its origin (on film or digital). If the image is initially in non-digital form, it is previously scanned.
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Color Television Image Signal Generators (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| FR0404837A FR2870071B1 (fr) | 2004-05-05 | 2004-05-05 | Procede de traitement de donnees d'images, par reduction de bruit d'image, et camera integrant des moyens de mise en oeuvre de ce procede |
| PCT/FR2005/000897 WO2005122086A2 (fr) | 2004-05-05 | 2005-04-13 | Reduction de bruit d’image numerique |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| EP1749278A2 true EP1749278A2 (fr) | 2007-02-07 |
Family
ID=34945477
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP05757055A Withdrawn EP1749278A2 (fr) | 2004-05-05 | 2005-04-13 | Procede de traitement de donnees d'image, par réduction de bruit d'image, et camera integrant des moyens de mise en oeuvre du procede |
Country Status (5)
| Country | Link |
|---|---|
| US (2) | US8253825B2 (fr) |
| EP (1) | EP1749278A2 (fr) |
| JP (1) | JP4999680B2 (fr) |
| FR (1) | FR2870071B1 (fr) |
| WO (1) | WO2005122086A2 (fr) |
Families Citing this family (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2870071B1 (fr) * | 2004-05-05 | 2006-12-08 | Centre Nat Rech Scient Cnrse | Procede de traitement de donnees d'images, par reduction de bruit d'image, et camera integrant des moyens de mise en oeuvre de ce procede |
| KR100791391B1 (ko) * | 2007-01-30 | 2008-01-07 | 삼성전자주식회사 | 노이즈 저감 방법 및 장치 |
| FR2919943B1 (fr) * | 2007-08-07 | 2010-02-26 | Dxo Labs | Procede de traitement d'objet numerique et systeme associe |
| FR2930360B1 (fr) * | 2008-04-21 | 2010-10-08 | Inst Nat Rech Inf Automat | Dispositif de traitement d'images ameliore. |
| US20100156768A1 (en) * | 2008-12-22 | 2010-06-24 | Fletcher Ii James Douglas | Display media, method of forming display media, and printer for printing on display media |
| JP4640508B2 (ja) | 2009-01-09 | 2011-03-02 | ソニー株式会社 | 画像処理装置、画像処理方法、プログラム、及び撮像装置 |
| US8331714B2 (en) | 2009-02-23 | 2012-12-11 | Sharp Laboratories Of America, Inc. | Methods and systems for image processing |
| JP5326920B2 (ja) * | 2009-08-07 | 2013-10-30 | 株式会社リコー | 画像処理装置、画像処理方法、及び、コンピュータプログラム |
| GB2481239B (en) | 2010-06-17 | 2014-04-02 | Canon Kk | Method and device for enhancing a digital image |
| JP6032467B2 (ja) * | 2012-06-18 | 2016-11-30 | 株式会社日立製作所 | 時空間データ管理システム、時空間データ管理方法、及びそのプログラム |
| KR101910870B1 (ko) * | 2012-06-29 | 2018-10-24 | 삼성전자 주식회사 | 잡음 제거 장치, 시스템 및 방법 |
| FR2994307B1 (fr) | 2012-08-06 | 2015-06-05 | Commissariat Energie Atomique | Procede et dispositif pour la reconstruction d'images a super-resolution |
| CN103686194B (zh) * | 2012-09-05 | 2017-05-24 | 北京大学 | 基于非局部均值的视频去噪方法和装置 |
| JP6071419B2 (ja) | 2012-10-25 | 2017-02-01 | キヤノン株式会社 | 画像処理装置及び画像処理方法 |
| JP6253331B2 (ja) * | 2012-10-25 | 2017-12-27 | キヤノン株式会社 | 画像処理装置及び画像処理方法 |
| TWI528815B (zh) * | 2013-02-04 | 2016-04-01 | 鈺立微電子股份有限公司 | 可降低影像雜訊之影像處理方法 |
| JP6161326B2 (ja) | 2013-02-27 | 2017-07-12 | キヤノン株式会社 | 画像処理装置、画像処理方法、及びプログラム |
| KR102144994B1 (ko) * | 2013-09-30 | 2020-08-14 | 삼성전자주식회사 | 영상의 노이즈를 저감하는 방법 및 이를 이용한 영상 처리 장치 |
| JP6332951B2 (ja) | 2013-11-29 | 2018-05-30 | キヤノン株式会社 | 画像処理装置および画像処理方法、およびプログラム |
| JP6501452B2 (ja) | 2014-04-04 | 2019-04-17 | キヤノン株式会社 | 画像処理装置及びシステム、画像処理方法、並びにプログラム |
| JP6465563B2 (ja) * | 2014-05-14 | 2019-02-06 | キヤノン株式会社 | 画像処理装置、画像処理方法およびプログラム |
| US9225889B1 (en) | 2014-08-18 | 2015-12-29 | Entropix, Inc. | Photographic image acquisition device and method |
| EP3132418B1 (fr) * | 2014-11-13 | 2019-10-09 | Huawei Technologies Co. Ltd. | Suppression non-locale du bruit d'image |
| JP6624785B2 (ja) * | 2015-01-16 | 2019-12-25 | キヤノン株式会社 | 画像処理方法、画像処理装置、撮像装置、プログラム、および、記憶媒体 |
| WO2016183743A1 (fr) | 2015-05-15 | 2016-11-24 | SZ DJI Technology Co., Ltd. | Système et procédé de prise en charge d'un débruitage d'image basé sur le degré de différentiation de blocs voisins |
| JP6614843B2 (ja) | 2015-07-30 | 2019-12-04 | キヤノン株式会社 | 画像処理装置、撮像装置および画像処理プログラム |
| UA115360C2 (uk) * | 2015-10-20 | 2017-10-25 | Максим Володимирович Давидов | Спосіб обробки цифрового зображення для зниження рівня шуму |
| US10282831B2 (en) * | 2015-12-28 | 2019-05-07 | Novatek Microelectronics Corp. | Method and apparatus for motion compensated noise reduction |
| US11750944B2 (en) | 2021-05-28 | 2023-09-05 | Varex Imaging Corporation | Pixel noise cancellation system |
| US11812187B2 (en) | 2021-05-28 | 2023-11-07 | Varex Imaging Corporation | Combined imaging array and strip |
| WO2022251716A1 (fr) * | 2021-05-28 | 2022-12-01 | Varex Imaging Corporation | Matrice et bande d'imagerie et système de suppression de bruit de pixel combinés |
| US11756285B2 (en) | 2021-06-10 | 2023-09-12 | Bank Of America Corporation | Image processing system and method for image noise removal |
Family Cites Families (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR970002967B1 (ko) * | 1993-04-09 | 1997-03-13 | 대우전자 주식회사 | 영역 분류패턴을 이용한 움직임벡터 검출장치 |
| US5461491A (en) * | 1993-12-29 | 1995-10-24 | Hewlett-Packard Company | Procedure for reducing processing time for image elements by combining charge of adjacent pixels into a new composite pixel |
| JP4056154B2 (ja) * | 1997-12-30 | 2008-03-05 | 三星電子株式会社 | 2次元連続映像の3次元映像変換装置及び方法並びに3次元映像の後処理方法 |
| KR100303086B1 (ko) | 1998-08-31 | 2001-09-24 | 전주범 | 적응적움직임추정장치 |
| JP2000293696A (ja) * | 1999-04-07 | 2000-10-20 | Matsushita Electric Ind Co Ltd | 画像認識装置 |
| JP2001094791A (ja) | 1999-09-27 | 2001-04-06 | Hitachi Kokusai Electric Inc | 画像ノイズ削減方法及び画像ノイズ削減装置並びに画像入力方法及び画像入力装置 |
| JP3576987B2 (ja) * | 2001-03-06 | 2004-10-13 | 株式会社東芝 | 画像のテンプレートマッチング方法及び画像処理装置 |
| KR100722596B1 (ko) * | 2001-04-19 | 2007-05-28 | 가부시끼가이샤 도시바 | 화상처리방법과 화상처리장치 |
| US7639741B1 (en) * | 2002-12-06 | 2009-12-29 | Altera Corporation | Temporal filtering using object motion estimation |
| FR2870071B1 (fr) * | 2004-05-05 | 2006-12-08 | Centre Nat Rech Scient Cnrse | Procede de traitement de donnees d'images, par reduction de bruit d'image, et camera integrant des moyens de mise en oeuvre de ce procede |
| JP2007135662A (ja) | 2005-11-15 | 2007-06-07 | Terumo Corp | 血液処理器具 |
| JP2008046746A (ja) | 2006-08-11 | 2008-02-28 | Olympus Corp | 工程管理システム |
-
2004
- 2004-05-05 FR FR0404837A patent/FR2870071B1/fr not_active Expired - Fee Related
-
2005
- 2005-04-13 US US11/579,380 patent/US8253825B2/en not_active Expired - Fee Related
- 2005-04-13 WO PCT/FR2005/000897 patent/WO2005122086A2/fr not_active Ceased
- 2005-04-13 EP EP05757055A patent/EP1749278A2/fr not_active Withdrawn
- 2005-04-13 JP JP2007512244A patent/JP4999680B2/ja not_active Expired - Fee Related
-
2012
- 2012-06-22 US US13/530,409 patent/US8427559B2/en not_active Expired - Fee Related
Non-Patent Citations (1)
| Title |
|---|
| None * |
Also Published As
| Publication number | Publication date |
|---|---|
| US8253825B2 (en) | 2012-08-28 |
| FR2870071A1 (fr) | 2005-11-11 |
| WO2005122086A2 (fr) | 2005-12-22 |
| JP2007536662A (ja) | 2007-12-13 |
| WO2005122086A3 (fr) | 2006-12-21 |
| US20100141804A1 (en) | 2010-06-10 |
| FR2870071B1 (fr) | 2006-12-08 |
| US8427559B2 (en) | 2013-04-23 |
| US20120268623A1 (en) | 2012-10-25 |
| JP4999680B2 (ja) | 2012-08-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP1749278A2 (fr) | Procede de traitement de donnees d'image, par réduction de bruit d'image, et camera integrant des moyens de mise en oeuvre du procede | |
| EP1410327B1 (fr) | Procede et systeme pour produire des informations formatees liees aux defauts d'au moins un appareil d'une chaine, notamment au flou | |
| JP5543605B2 (ja) | 空間画像事前確率を用いたぼけ画像修正 | |
| FR2851676A1 (fr) | Procede d'obtention d'une image a resolution augmentee par l'utilisation d'une pluralite d'images a faible resolution | |
| WO2011010040A1 (fr) | Procede d'estimation d'un defaut d'un systeme de capture d'images et systemes associes | |
| Al-Hatmi et al. | A review of Image Enhancement Systems and a case study of Salt &pepper noise removing | |
| Mosleh et al. | Explicit ringing removal in image deblurring | |
| US7602989B2 (en) | Realtime 2D deconvolution system and method | |
| Mishra et al. | Digital image restoration using image filtering techniques | |
| FR3088755A1 (fr) | Procede de defloutage d’une image | |
| Narasimharao et al. | Advanced techniques for color image blind deconvolution to restore blurred images | |
| Khan et al. | Digital image restoration in matlab: A case study on inverse and wiener filtering | |
| Chittilappilly et al. | SVM based defect detection for industrial applications | |
| Safonov et al. | Adaptive sharpening of photos | |
| Ahalya et al. | Deep Learning for Single Image Deblurring | |
| Chahal et al. | Robust statistics based filter to remove salt and pepper noise in digital images | |
| Chopra et al. | Low light GAN-based photo enhancement | |
| Bundela et al. | Restoration of Single Blur Image Using Blind Deconvolution Method | |
| Lata et al. | Novel method to assess motion blur kernel parameters and comparative study of restoration techniques using different image layouts | |
| Hulaj et al. | Removal of various noises from digital images with the application of a single algorithm | |
| FR3066633A1 (fr) | Procede de defloutage d'une image | |
| Lisha et al. | Single image super resolution of blurred natural images using blur kernel estimation combined with super resolution convolution neural network | |
| Kazi et al. | Image Restoration using Blind Deconvolution | |
| Dinesh et al. | Image Processing | |
| CN121437354A (zh) | 图像的检测方法 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
| 17P | Request for examination filed |
Effective date: 20061106 |
|
| AK | Designated contracting states |
Kind code of ref document: A2 Designated state(s): AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LI LT LU MC NL PL PT RO SE SI SK TR |
|
| AX | Request for extension of the european patent |
Extension state: AL BA HR LV MK YU |
|
| DAX | Request for extension of the european patent (deleted) | ||
| 17Q | First examination report despatched |
Effective date: 20070823 |
|
| GRAP | Despatch of communication of intention to grant a patent |
Free format text: ORIGINAL CODE: EPIDOSNIGR1 |
|
| INTG | Intention to grant announced |
Effective date: 20180727 |
|
| RIN1 | Information on inventor provided before grant (corrected) |
Inventor name: MOREL, JEAN-MICHEL Inventor name: BUADES, ANTONIO Inventor name: COLL, BARTOMEU |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN |
|
| 18D | Application deemed to be withdrawn |
Effective date: 20181207 |